Rewrite the way GIN posting lists are packed on a page, to reduce WAL volume.

Inserting (in retail) into the new 9.4 format GIN posting tree created much
larger WAL records than in 9.3. The previous strategy to WAL logging was
basically to log the whole page on each change, with the exception of
completely unmodified segments up to the first modified one. That was not
too bad when appending to the end of the page, as only the last segment had
to be WAL-logged, but per Fujii Masao's testing, even that produced 2x the
WAL volume that 9.3 did.

The new strategy is to keep track of changes to the posting lists in a more
fine-grained fashion, and also make the repacking" code smarter to avoid
decoding and re-encoding segments unnecessarily.

src/backend/access/gin/ginpostinglist.c

@@ -298,9 +298,10 @@ ginPostingListDecodeAllSegments(GinPostingList *segment, int len, int *ndecoded_
 		}
 
 		/* copy the first item */
+		Assert(OffsetNumberIsValid(ItemPointerGetOffsetNumber(&segment->first)));
+		Assert(ndecoded == 0 || ginCompareItemPointers(&segment->first, &result[ndecoded - 1]) > 0);
 		result[ndecoded] = segment->first;
 		ndecoded++;
-		Assert(OffsetNumberIsValid(ItemPointerGetOffsetNumber(&segment->first)));
 
 		val = itemptr_to_uint64(&segment->first);
 		ptr = segment->bytes;

src/backend/access/gin/ginxlog.c

@@ -145,15 +145,158 @@ ginRedoInsertEntry(Buffer buffer, bool isLeaf, BlockNumber rightblkno, void *rda
 static void
 ginRedoRecompress(Page page, ginxlogRecompressDataLeaf *data)
 {
-	Pointer		segment;
-
-	/* Copy the new data to the right place */
-	segment = ((Pointer) GinDataLeafPageGetPostingList(page))
-		+ data->unmodifiedsize;
-	memcpy(segment, data->newdata, data->length - data->unmodifiedsize);
-	GinDataLeafPageSetPostingListSize(page, data->length);
-	GinPageSetCompressed(page);
-	GinPageGetOpaque(page)->maxoff = InvalidOffsetNumber;
+	int			actionno;
+	int			segno;
+	GinPostingList *oldseg;
+	Pointer		segmentend;
+	char	   *walbuf;
+	int			totalsize;
+
+	/*
+	 * If the page is in pre-9.4 format, convert to new format first.
+	 */
+	if (!GinPageIsCompressed(page))
+	{
+		ItemPointer uncompressed = (ItemPointer) GinDataPageGetData(page);
+		int			nuncompressed = GinPageGetOpaque(page)->maxoff;
+		int			npacked;
+		GinPostingList *plist;
+
+		plist = ginCompressPostingList(uncompressed, nuncompressed,
+									   BLCKSZ, &npacked);
+		Assert(npacked == nuncompressed);
+
+		totalsize = SizeOfGinPostingList(plist);
+
+		memcpy(GinDataLeafPageGetPostingList(page), plist, totalsize);
+		GinDataLeafPageSetPostingListSize(page, totalsize);
+		GinPageSetCompressed(page);
+		GinPageGetOpaque(page)->maxoff = InvalidOffsetNumber;
+	}
+
+	oldseg = GinDataLeafPageGetPostingList(page);
+	segmentend = (Pointer) oldseg + GinDataLeafPageGetPostingListSize(page);
+	segno = 0;
+
+	walbuf = ((char *) data) + sizeof(ginxlogRecompressDataLeaf);
+	for (actionno = 0; actionno < data->nactions; actionno++)
+	{
+		uint8		a_segno = *((uint8 *) (walbuf++));
+		uint8		a_action = *((uint8 *) (walbuf++));
+		GinPostingList *newseg = NULL;
+		int			newsegsize = 0;
+		ItemPointerData *items = NULL;
+		uint16		nitems = 0;
+		ItemPointerData *olditems;
+		int			nolditems;
+		ItemPointerData *newitems;
+		int			nnewitems;
+		int			segsize;
+		Pointer		segptr;
+		int			szleft;
+
+		/* Extract all the information we need from the WAL record */
+		if (a_action == GIN_SEGMENT_INSERT ||
+			a_action == GIN_SEGMENT_REPLACE)
+		{
+			newseg = (GinPostingList *) walbuf;
+			newsegsize = SizeOfGinPostingList(newseg);
+			walbuf += SHORTALIGN(newsegsize);
+		}
+
+		if (a_action == GIN_SEGMENT_ADDITEMS)
+		{
+			memcpy(&nitems, walbuf, sizeof(uint16));
+			walbuf += sizeof(uint16);
+			items = (ItemPointerData *) walbuf;
+			walbuf += nitems * sizeof(ItemPointerData);
+		}
+
+		/* Skip to the segment that this action concerns */
+		Assert(segno <= a_segno);
+		while (segno < a_segno)
+		{
+			oldseg = GinNextPostingListSegment(oldseg);
+			segno++;
+		}
+
+		/*
+		 * ADDITEMS action is handled like REPLACE, but the new segment to
+		 * replace the old one is reconstructed using the old segment from
+		 * disk and the new items from the WAL record.
+		 */
+		if (a_action == GIN_SEGMENT_ADDITEMS)
+		{
+			int			npacked;
+
+			olditems = ginPostingListDecode(oldseg, &nolditems);
+
+			newitems = ginMergeItemPointers(items, nitems,
+											olditems, nolditems,
+											&nnewitems);
+			Assert(nnewitems == nolditems + nitems);
+
+			newseg = ginCompressPostingList(newitems, nnewitems,
+											BLCKSZ, &npacked);
+			Assert(npacked == nnewitems);
+
+			newsegsize = SizeOfGinPostingList(newseg);
+			a_action = GIN_SEGMENT_REPLACE;
+		}
+
+		segptr = (Pointer) oldseg;
+		if (segptr != segmentend)
+			segsize = SizeOfGinPostingList(oldseg);
+		else
+		{
+			/*
+			 * Positioned after the last existing segment. Only INSERTs
+			 * expected here.
+			 */
+			Assert(a_action == GIN_SEGMENT_INSERT);
+			segsize = 0;
+		}
+		szleft = segmentend - segptr;
+
+		switch (a_action)
+		{
+			case GIN_SEGMENT_DELETE:
+				memmove(segptr, segptr + segsize, szleft - segsize);
+				segmentend -= segsize;
+
+				segno++;
+				break;
+
+			case GIN_SEGMENT_INSERT:
+				/* make room for the new segment */
+				memmove(segptr + newsegsize, segptr, szleft);
+				/* copy the new segment in place */
+				memcpy(segptr, newseg, newsegsize);
+				segmentend += newsegsize;
+				segptr += newsegsize;
+				break;
+
+			case GIN_SEGMENT_REPLACE:
+				/* shift the segments that follow */
+				memmove(segptr + newsegsize,
+						segptr + segsize,
+						szleft - segsize);
+				/* copy the replacement segment in place */
+				memcpy(segptr, newseg, newsegsize);
+				segmentend -= segsize;
+				segmentend += newsegsize;
+				segptr += newsegsize;
+				segno++;
+				break;
+
+			default:
+				elog(ERROR, "unexpected GIN leaf action: %u", a_action);
+		}
+		oldseg = (GinPostingList *) segptr;
+	}
+
+	totalsize = segmentend - (Pointer) GinDataLeafPageGetPostingList(page);
+	GinDataLeafPageSetPostingListSize(page, totalsize);
 }
 
 static void

src/backend/access/rmgrdesc/gindesc.c

@@ -25,6 +25,57 @@ desc_node(StringInfo buf, RelFileNode node, BlockNumber blkno)
 					 node.spcNode, node.dbNode, node.relNode, blkno);
 }
 
+static void
+desc_recompress_leaf(StringInfo buf, ginxlogRecompressDataLeaf *insertData)
+{
+	int			i;
+	char	   *walbuf = ((char *) insertData) + sizeof(ginxlogRecompressDataLeaf);
+
+	appendStringInfo(buf, " %d segments:", (int) insertData->nactions);
+
+	for (i = 0; i < insertData->nactions; i++)
+	{
+		uint8		a_segno = *((uint8 *) (walbuf++));
+		uint8		a_action = *((uint8 *) (walbuf++));
+		uint16		nitems = 0;
+		int			newsegsize = 0;
+
+		if (a_action == GIN_SEGMENT_INSERT ||
+			a_action == GIN_SEGMENT_REPLACE)
+		{
+			newsegsize = SizeOfGinPostingList((GinPostingList *) walbuf);
+			walbuf += SHORTALIGN(newsegsize);
+		}
+
+		if (a_action == GIN_SEGMENT_ADDITEMS)
+		{
+			memcpy(&nitems, walbuf, sizeof(uint16));
+			walbuf += sizeof(uint16);
+			walbuf += nitems * sizeof(ItemPointerData);
+		}
+
+		switch(a_action)
+		{
+			case GIN_SEGMENT_ADDITEMS:
+				appendStringInfo(buf, " %d (add %d items)", a_segno, nitems);
+				break;
+			case GIN_SEGMENT_DELETE:
+				appendStringInfo(buf, " %d (delete)", a_segno);
+				break;
+			case GIN_SEGMENT_INSERT:
+				appendStringInfo(buf, " %d (insert)", a_segno);
+				break;
+			case GIN_SEGMENT_REPLACE:
+				appendStringInfo(buf, " %d (replace)", a_segno);
+				break;
+			default:
+				appendStringInfo(buf, " %d unknown action %d ???", a_segno, a_action);
+				/* cannot decode unrecognized actions further */
+				return;
+		}
+	}
+}
+
 void
 gin_desc(StringInfo buf, uint8 xl_info, char *rec)
 {
@@ -70,9 +121,10 @@ gin_desc(StringInfo buf, uint8 xl_info, char *rec)
 					ginxlogRecompressDataLeaf *insertData =
 						(ginxlogRecompressDataLeaf *) payload;
 
-					appendStringInfo(buf, " unmodified: %u length: %u (compressed)",
-									 insertData->unmodifiedsize,
-									 insertData->length);
+					if (xl_info & XLR_BKP_BLOCK(0))
+						appendStringInfo(buf, " (full page image)");
+					else
+						desc_recompress_leaf(buf, insertData);
 				}
 				else
 				{
@@ -105,9 +157,10 @@ gin_desc(StringInfo buf, uint8 xl_info, char *rec)
 				ginxlogVacuumDataLeafPage *xlrec = (ginxlogVacuumDataLeafPage *) rec;
 				appendStringInfoString(buf, "Vacuum data leaf page, ");
 				desc_node(buf, xlrec->node, xlrec->blkno);
-				appendStringInfo(buf, " unmodified: %u length: %u",
-							 xlrec->data.unmodifiedsize,
-							 xlrec->data.length);
+				if (xl_info & XLR_BKP_BLOCK(0))
+					appendStringInfo(buf, " (full page image)");
+				else
+					desc_recompress_leaf(buf, &xlrec->data);
 			}
 			break;
 		case XLOG_GIN_DELETE_PAGE:

src/include/access/gin_private.h

@@ -406,7 +406,8 @@ typedef struct
 	 * whose split this insertion finishes. As BlockIdData[2] (beware of adding
 	 * fields before this that would make them not 16-bit aligned)
 	 *
-	 * 2. one of the following structs, depending on tree type.
+	 * 2. an ginxlogInsertEntry or ginxlogRecompressDataLeaf struct, depending
+	 * on tree type.
 	 *
 	 * NB: the below structs are only 16-bit aligned when appended to a
 	 * ginxlogInsert struct! Beware of adding fields to them that require
@@ -421,15 +422,39 @@ typedef struct
 	IndexTupleData tuple;	/* variable length */
 } ginxlogInsertEntry;
 
+
 typedef struct
 {
-	uint16		length;
-	uint16		unmodifiedsize;
+	uint16		nactions;
 
-	/* compressed segments, variable length */
-	char		newdata[1];
+	/* Variable number of 'actions' follow */
 } ginxlogRecompressDataLeaf;
 
+/*
+ * Note: this struct is currently not used in code, and only acts as
+ * documentation. The WAL record format is as specified here, but the code
+ * uses straight access through a Pointer and memcpy to read/write these.
+ */
+typedef struct
+{
+	uint8		segno;		/* segment this action applies to */
+	char		type;		/* action type (see below) */
+
+	/*
+	 * Action-specific data follows. For INSERT and REPLACE actions that is a
+	 * GinPostingList struct. For ADDITEMS, a uint16 for the number of items
+	 * added, followed by the items themselves as ItemPointers. DELETE actions
+	 * have no further data.
+	 */
+} ginxlogSegmentAction;
+
+/* Action types */
+#define GIN_SEGMENT_UNMODIFIED	0	/* no action (not used in WAL records) */
+#define GIN_SEGMENT_DELETE		1	/* a whole segment is removed */
+#define GIN_SEGMENT_INSERT		2	/* a whole segment is added */
+#define GIN_SEGMENT_REPLACE		3	/* a segment is replaced */
+#define GIN_SEGMENT_ADDITEMS	4	/* items are added to existing segment */
+
 typedef struct
 {
 	OffsetNumber offset;